Remaining Challenges in Assessing Non-Inferiority.

Despite decades of experience with non-inferiority trials, they remain a source of great controversy and involve fundamental areas of disagreement. Indeed, there can be little more fundamental than the definition of the hypothesis to be tested, and yet it can be argued that the null hypothesis commonly used in these trials, which involves specification of a non-inferiority margin, is the wrong one. Non-inferiority trials involve an indirect comparison between the experimental treatment under evaluation and placebo, but the null hypothesis based on a non-inferiority margin does not address this aspect of the trial's objective. In addition, the criterion in common use for concluding that the experimental treatment preserves a prespecified fraction of the control treatment's effect is in conflict with a reasonable criterion for preference of the experimental treatment over the control, and can lead to inconsistencies in drug approval decisions. This review discusses these issues and provides recommendations.